"We are focusing on the foundation to achieve better parallelism, better scalability, eliminate all latency bottlenecks and improve real-time responsiveness"
As processing humongous video content ‘manually’ is next to impossible, AvidBeam unlocks the power of big data by leveraging machine learning as a fundamental tool that delivers deep learning techniques to the detection and recognition process coupled, with significant breakthroughs in accuracy improvements. At the core, AvidBeam leverages three main infrastructure components to build products targeting surveillance and security, retail, automotive, industrial and consumer space. The first infrastructure component is ATUN™– an open, extensible platform built using big data tools along with proprietary schedulers and resource managers. Leveraging ATUN™, the company can scale video processing linearly with the number of available resources to reach very high processing rate. The second key component of the infrastructure is a library for computer vision algorithms—ViBE™ that helps in detecting and recognizing sequential and non-sequential types of events and applying deep learning models for detecting and recognition of objects. This computer vision algorithm constitutes map functions that are plugged into ATUN™ to accelerate the rate of scaling the video processing.
Further, with the help of third infrastructure component DiVA™, the company can aggregate the video streams from the edge and consolidate on the cloud for processing. It also exercises bandwidth savings techniques over video streams to make efficient use of the available link size. AvidBeam also has built multiple applications on top of these components all running on the cloud taking advantage of big data technology to scale video processing.
Our goal is to deliver computer vision services to the customers in all sectors in the most scalable way and using the most feasible amount of resources
“All of these solutions make effective use of big data and machine learning to achieve scalability while improving accuracy over time,” asserts El Gebaly.
AvidBeam’s video analytic platform can process a massive amount of media files and improve computer vision accuracy over time by applying big data and deep learning techniques. Further, AvidBeam’s prowess in machine learning, big data, visualization and computer vision technologies help them to deliver custom solutions to interested parties per requested specification.
In terms of targeted market segment, the company’s prime focus is on two main markets: surveillance and automotive. As the pull for video analytics in these markets is growing, AvidBeam’s video analytic platform helps in automatic tagging and annotation of video, along with LiDAR systems that can capture scenes and save tremendous time and effort.
To elaborate more on the company’s products and their value proposition, El Gebaly cites a scenario, where a top camera surveillance vendor in Shanghai, was looking for a cutting-edge technology to scale video processing of live camera feeds for thousands of cameras deployed in the field. By leveraging AvidBeam’s cloud-based video analytics platform, the camera surveillance vendor was able to process a large number of input camera feeds simultaneously and in real-time. AvidBeam also derived a formula that described the relationship between real-time processing and number of available resources for a given set of input feeds. While scripting similar success stories for a plethora of clients, the company is addressing the significant need for real-time processing with its robust video analytics platform. “We aim to be the number one scalable video processing engine in the world,” says the CEO.
Along the path of innovation, AvidBeam looks forward to developing additional applications in various key segments on top of ATUN and ViBE while keeping the total cost of ownership low for the customers. The company also foresees tackling new verticals in the near future primarily medicine and entertainment where ATUN™ and ViBE™ play a significant role in offering new information services to members of the community. “Finally, we are focusing on the foundation to achieve better parallelism, better scalability, eliminate all latency bottlenecks and improve real-time responsiveness,” concludes El Gebaly.